3DGS stands for 3D Gaussian Splat. This is a method for capturing a scene or object by utilizing photos or video that are processed into a 3D environment by matching reference points shared between images. When photos or video are captured of the same object, from multiple angles, it allows for the generation of something like a 3D point cloud, but in this case the point cloud is made of fine strands of color data that are overlapped (splat). This method of reality capture is nice because it often creates a more realistic looking scene or object than photogrammetry, especially when fine details like the grass in the scene below need to be visualized.
The video below shows some of the capabilities of this technology. The first portion of the video shows some of the base information from the source images that are used for 3D reconstruction and it therefore looks more like a point cloud. The second portion shows a depth map that can be used to integrate other 3D objects into the scene or the depth map can be used to adjust the focus of the 3D camera, making foreground or background portions of the scene more in-focus or blurry.
What is intriguing to me about this technology is the ability to capture a point in time which can then be reconstructed and experienced at from any point of view. This allows for camera work to be handled in post-production, for example. I'm looking forward to seeing where this technology goes.
This video shows some of a process for making the 3DGS above.